Lunar Crater Identification in Digital Images

Area of Science:

• Planetary Science
• Computer Vision
• Robotics

Background:

• Crater identification in lunar images is crucial for terrain relative navigation (TRN) and image registration.
• Previous methods relied on heuristic schemes, limiting performance in general scenarios.
• The

Purpose of the Study:

• To develop a mathematically rigorous solution for the general crater identification problem in lunar imagery.
• To enable accurate crater pattern recognition without prior camera pose information.
• To advance the capabilities of crater-based TRN systems.

Main Methods:

• Developed a mathematically rigorous framework for analyzing elliptical crater rims under perspective projection.
• Utilized invariant theory to create viewpoint-invariant descriptors for crater patterns.
• Introduced novel techniques for pose computation and crater rim correspondence evaluation.

Main Results:

• Established conditions for the recognizability of crater patterns in perspective images.
• Demonstrated the effectiveness of invariant descriptors for robust crater identification.
• Validated new pose estimation and correspondence techniques on synthetic and real lunar images.

Conclusions:

• The proposed invariant-theory-based approach provides a provably robust solution to the general crater identification problem.
• This work significantly improves upon heuristic methods, enabling reliable navigation and registration in diverse imaging conditions.
• The developed techniques offer a foundation for enhanced autonomous operations in lunar exploration.